CN113688759A - Safety helmet identification method based on deep learning - Google Patents
Safety helmet identification method based on deep learning Download PDFInfo
- Publication number
- CN113688759A CN113688759A CN202111008443.1A CN202111008443A CN113688759A CN 113688759 A CN113688759 A CN 113688759A CN 202111008443 A CN202111008443 A CN 202111008443A CN 113688759 A CN113688759 A CN 113688759A
- Authority
- CN
- China
- Prior art keywords
- safety helmet
- person
- frame
- model
- helmet
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000013135 deep learning Methods 0.000 title claims abstract description 7
- 238000010276 construction Methods 0.000 claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 9
- 238000010586 diagram Methods 0.000 claims description 9
- 230000006872 improvement Effects 0.000 claims description 9
- 230000007246 mechanism Effects 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000004088 simulation Methods 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 claims description 2
- 238000013136 deep learning model Methods 0.000 claims 1
- 238000002372 labelling Methods 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 230000032683 aging Effects 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Helmets And Other Head Coverings (AREA)
Abstract
The invention provides a safety helmet identification method based on deep learning, which comprises the steps of firstly, acquiring image data of safety helmets worn by personnel in various construction sites; the images are clearer and richer and have certain generalization capability by adopting a plurality of data enhancement methods, and the images subjected to data enhancement are made into a safety helmet data set; the native YOLOv5 model is improved, so that the detection accuracy and the detection speed of the safety helmet can be improved; after the model training is finished, the model is used for detecting the video stream of a construction site, and secondary coincidence matching is carried out on the detected target frame so as to improve the detection accuracy and reduce the false alarm rate in actual use; and timely warning the system when no personnel wearing the safety helmet is present.
Description
Technical Field
The invention belongs to the field of artificial intelligence, and particularly relates to an improved helmet identification method based on YOLOv 5.
Background
In construction sites of construction works, safety is the most important and essential requirement for workers. Because production and construction operation sites are numerous and widely distributed, the site environment is very complex, construction operators are numerous, and workers in charge of supervision are limited, the comprehensive, whole-process and real-time safety supervision and management of the production and construction operation sites are difficult to achieve. Although a great deal of monitoring equipment is installed on many production and construction operation sites at present, supervisors can check the monitoring materials after a period of time, the requirements of safety monitoring on the construction site cannot be met for accidents or unsafe behaviors which have already occurred, and the personal loss and property loss caused by the accidents cannot be recovered.
In the construction operation field, a large amount of safety appliances are needed, and the safety helmet is an important safety appliance, plays a role in protecting the head and prevents the head from being injured by falling objects and other factors. Most production and construction sites at present have the regulations of the safety helmet being worn correctly, but still have a lot of behaviors of not wearing the safety helmet. In order to effectively improve the situation and reduce the occurrence of similar behaviors, the modern computer vision technology with real-time performance, high precision and robustness is necessary to detect whether the personnel on the construction site wear the safety helmet or not.
The safety helmet detection system is improved based on YOLOv5, mainly aims at whether a worker in a construction site wears a safety helmet or not, can detect the worker in a range in real time through a camera, judge whether the worker wears the safety helmet or not, and immediately send out an alarm through an alarm system if the worker who does not wear the safety helmet is found, record the alarm and provide reliable information for background monitoring personnel. The method improves the safety of the production environment and can greatly improve the guarantee of the safety production of enterprises.
Disclosure of Invention
The invention provides an improved safety helmet identification method based on YOLOv5 based on the existing construction site accidents caused by frequent wearing of safety helmets, and a target frame matching method is added on the basis of a general computer vision detection method, so that the detection precision of wearing of safety helmets by personnel is improved.
The invention specifically comprises the following contents: an improved safety helmet identification method based on YOLOv5 comprises the following five steps of data acquisition, data set making and enhancing, Yolov5 model improving and training, safety helmet wearing detection and decision alarming:
step S1: collecting image data of safety helmets worn by personnel in various construction sites;
step S2: carrying out data set division and personnel safety helmet wearing target frame marking on the collected personnel safety helmet image data; and after the annotation is finished, performing data enhancement on the original image data set, wherein the image enhancement method comprises the following steps: self-adaptive Gaussian filtering denoising, noise simulation of rain, snow and fog weather conditions;
step S3: carrying out wearing detection improvement aiming at the personnel safety helmet on a YOLOv5 model, wherein the wearing detection improvement comprises adding a ghost bottleeck detection mechanism and an attention mechanism to a backbone network; training the model by using the manufactured data set after the improvement is completed, and verifying the detection speed and the detection precision of the model;
step S4: the trained YOLOv5 model is used for detecting the input construction site video stream in real time and judging whether a construction site person wears a safety helmet or not;
step S5: if the model judges that a person does not wear the safety helmet, recording the moment and storing video streams of a period of time before and after the moment;
in order to better implement the present invention, further, the step S1 specifically includes the following steps: erect camera of different angles, height at all kinds of job sites, set up the parameter of the video of camera shooting into resolution ratio: 1920 × 1080, frame rate: and 30, continuously shooting the wearing condition of the safety helmet of the personnel in the construction site. When more videos are obtained, videos with better angles and clear personnel are screened out and frames are randomly extracted from the videos, and the extracted pictures are used as image data of the safety helmet worn by the personnel.
In order to better implement the present invention, further, the step S2 specifically includes the following steps:
step S2.1: marking a target frame and making a data set, marking the target frame on image data of a safety helmet worn by a person by using a labellmg tool, wherein the types of the target frame are as follows: person, head, helmet. And then dividing the annotated image into a training set, a testing set and a verification set according to the ratio of 6:2: 2.
Step S2.2: and (3) carrying out self-adaptive Gaussian filtering denoising on the data set manufactured in the step (S2.1), wherein the image blur can occur in the image acquired in the actual construction site due to poor quality or aging of the camera, and the Gaussian filtering denoising on the image can improve the image details and make the image clearer. The two-dimensional gaussian function formula used is as follows:
wherein, σ is the standard deviation of normal distribution, and the value thereof determines the variation amplitude of the Gaussian function, and corresponds to the weight of the filter.
Step S2.3: : simulating rain and snow weather and fog weather conditions. In an actual construction site, various weather conditions cannot be completely covered due to limited shooting time, and the data can be richer by simulating the various weather conditions. The specific method comprises the following steps: the method comprises the steps of using OPENCV to generate random noise with different densities to simulate rain, snow and fog under different conditions, then stretching and rotating the random noise, simulating the rain, snow and fog in the direction, and superposing the noise and an original image to obtain a simulated rain, snow and fog weather fieldA scene; the mathematical formula of the randomly generated noise is as follows:
given an initial value x0Generating a random number y over the (0,1) intervali. Wherein: 2000 for a, 1 for c, M225(ii) a Then by transforming Zi＝a+(b-a)yiGenerating a random number Z over the (a, b) intervali。
In order to better implement the present invention, further, the step S3 specifically includes the following steps:
step S3.1: a ghestbottleeck detection mechanism is added into native YOLOv5, namely a GHOST module is used in a Neck network of YOLOv5, the total number of required parameters is reduced under the condition that the size of an output feature map is not changed, and the calculation complexity is reduced.
Step S3.2: ghostnet is a novel network structure, and on the basis of a small amount of feature maps obtained by nonlinear convolution, linear convolution is performed once, so that more feature maps are obtained, redundant features are eliminated, and a lighter model is obtained. In the improvement of the native YOLOv5, a module ghostbotleneck with a Ghostnet network structure is used for replacing a bottleneckcps structure in the YOLOv5, so that the computation parameters of the native YOLOv5 are reduced, and the training speed of the model can be improved.
Step S3.3: after the improvement of the backbone network and the NECK part of YOLOv5 is completed, the improved model is trained by using the data set prepared in step S2.3, and the model weight is saved after the model is trained.
In order to better implement the present invention, further, the step S4 specifically includes the following steps:
step S4.1: and (4) applying the model trained in the step (S3.3) to safety helmet detection on a construction site, wherein 3 types of target frames person, head and helmet detected by the model carry out the following steps: when the person frame appears, whether the head frame and the helmet frame appear in the person frame or not is continuously detected, and if the person frame does not appear, the head frame and the helmet frame are not detected. When the person detects the person frame and the helmet frame, the person can match the two frames according to the contact ratio, and when the matching ratio is higher than 13%, the person can be judged to wear the safety helmet. And if the person frame and the head frame are detected, and the matching degree of the person frame and the head frame is higher than 13%, the user is judged that the safety helmet is not worn.
Step S4.2: and detecting the presence of a person who does not wear the safety helmet, and sending alarm information to the system.
Compared with the prior art, the invention has the following advantages:
1): the hardware requirements on the detection end and the data acquisition end are not high;
2): the backbone network of YOLOv5 is improved, an SE module is added in a Focus structure, and a small number of calculation modules are added to improve the detection accuracy of the model;
3): the Neck layer of YOLOv5 is improved, and the detection speed of the model is improved under the condition of no performance loss;
4): and target frame coincidence degree detection is added on the basis of model detection, so that the false alarm rate is reduced in the actual application.
Drawings
FIG. 1 data enhancement schematic
FIG. 2 is a schematic diagram of a ghestbottleeck module
FIG. 3 SE Module schematic
FIG. 4 is a modified Yolov5 model structure diagram
FIG. 5 is a diagram illustrating a wearing judgment of a helmet
The specific implementation mode is as follows:
in order to more clearly illustrate the technical solutions of the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it should be understood that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments, and therefore should not be considered as a limitation to the scope of protection. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
An improved safety cap identification method based on YOLOv5 is disclosed, and in combination with the steps of FIG. 1, FIG. 2, FIG. 3, FIG. 4 and FIG. 5, the method comprises the following five steps of data acquisition, data enhancement, improved Yolov5 model training, safety cap wearing detection and decision alarm:
step S1: collecting image data of safety helmets worn by personnel in various construction sites; erect camera of different angles, height at all kinds of job sites, set up the parameter of the video of camera shooting into resolution ratio: 1920 × 1080, frame rate: and 30, continuously shooting the wearing condition of the safety helmet of the personnel in the construction site. When more videos are obtained, videos with better angles and clear personnel are screened out and frames are randomly extracted from the videos, and the extracted pictures are used as image data of the safety helmet worn by the personnel.
Step S2: and (3) marking the acquired number of images of the safety helmet worn by the person by using a labellmg tool, wherein the types of the target frames are as follows: person, head, helmet. And then dividing the annotated image into a training set, a testing set and a verification set according to the ratio of 6:2: 2.
After the data set is divided, enhancing the image data of the safety helmet worn by the person: the image blur can appear in the image that acquires in the actual building site because of camera quality is poor or ageing, carries out the Gaussian filter to remove the noise and can promote the picture detail to the picture, makes the picture clearer. The two-dimensional gaussian function formula used is as follows:
wherein, σ is the standard deviation of normal distribution, and the value thereof determines the variation amplitude of the Gaussian function, and corresponds to the weight of the filter.
In an actual construction site, various weather conditions cannot be completely covered due to limited shooting time, and the data can be richer by simulating the various weather conditions. The specific method comprises the following steps: the method comprises the steps of generating random noise with different densities by using an OPENCV to simulate rain, snow and fog with different sizes, setting uniform random numbers and threshold values to control the level of the noise, controlling the elongation, the rotation direction and the allowance of the noise to simulate the rain, snow and fog in different directions, and superposing the generated noise and an original picture to obtain a simulated rain, snow and fog weather picture, wherein the mathematical formula for randomly generating the noise is as follows:
given an initial value x0Generating a random number y over the (0,1) intervali. Wherein: 2000 for a, 1 for c, M225(ii) a Then by transforming Zi＝a+(b-a)yiGenerating a random number Z over the (a, b) intervali. The obtained simulated picture is shown as a data enhancement schematic diagram in fig. 1.
Step S3: the YOLOv5 model is improved in a targeted mode, and a Neck layer is improved: a module ghostbottleck with a Ghostnet network structure is used for replacing a bottleckcps structure in YOLOv5, as shown in a schematic diagram of a ghostbottleck in figure 2, so as to reduce the calculation parameters of native YOLOv 5. Improving a backbone network: an SElayer module is added into a Focus structure to optimize the learned content and improve the sensitivity of the model to channel characteristics, and a SE module schematic diagram is shown in FIG. 3.
After the improvement of the backbone network and the NECK part of YOLOv5 is completed, the improved structure diagram of the YOOLOv5 network is shown in fig. 4, the improved model is trained by using the data set created in step S2.3, and the model weight is saved after the model is trained.
Step S4: the model trained in the step S3 is used for detecting a safety helmet on a construction site, and the 3 types of target frames person, head, and helmet detected by the model perform the following steps: when the person frame appears, whether the head frame and the helmet frame appear in the person frame or not is continuously detected, and if the person frame does not appear, the head frame and the helmet frame are not detected. When the person detects the person frame and the helmet frame, the person can match the two frames according to the contact ratio, and when the matching ratio is higher than 13%, the person can be judged to wear the safety helmet. If the person frame and the head frame are detected, and the matching degree of the person frame and the head frame is higher than 13%, the safety helmet is not worn, and the judgment process is shown in a safety helmet wearing judgment diagram in fig. 5.
Step S5: if the model judges that a person does not wear the safety helmet, recording the moment and storing video streams of a period of time before and after the moment;
step S4.2: and detecting the presence of a person who does not wear the safety helmet, and sending alarm information to the system.
Claims (4)
1. A safety helmet identification method based on deep learning is characterized by comprising the following steps:
step S1: collecting image data of safety helmets worn by personnel in various construction sites;
step S2: carrying out data set division and personnel safety helmet wearing target frame marking on the collected personnel safety helmet image data; and after the annotation is finished, performing data enhancement on the original image data set, wherein the image enhancement method comprises the following steps: self-adaptive Gaussian filtering denoising, noise simulation of rain, snow and fog weather conditions;
step S3: carrying out wearing detection improvement aiming at the personnel safety helmet on a deep learning model YOLOv5, wherein the wearing detection improvement comprises adding a ghestbottleneck detection mechanism and an attention mechanism to a backbone network; training the model by using the manufactured data set after the improvement is completed, and verifying the detection speed and the detection precision of the model;
step S4: the trained improved YOLOv5 model is used for detecting the input construction site video stream in real time and judging whether a construction site person wears a safety helmet or not;
step S5: and if the model judges that the person does not wear the safety helmet, recording the moment and storing the video stream a period of time before and after the moment.
2. The safety helmet identification method based on deep learning of claim 1 is mainly characterized in that: the step S2 includes the steps of:
step S2.1: carrying out personnel safety helmet wearing target frame labeling on the obtained images of various construction sites, wherein the types of the target frames are as follows: person, head, helmet;
step S2.2: carrying out Gaussian filtering denoising on the marked personnel safety helmet wearing image, improving the definition of image data and amplifying picture details;
step S2.3: to the obtained part of construction siteNoise simulation is carried out on the interpatient images to simulate various weather conditions, the simulation is mainly carried out on rainy and snowy days, foggy days and dusty days, random noises with different densities are generated by using an OPENCV to simulate the rainy, snowy and foggy days under different conditions, then the random noises are lengthened and rotated in directions to simulate the rainy, snowy and foggy days in the directions, and the noises and the original images are superposed to obtain a simulated rainy, snowy and foggy weather scene; the mathematical formula of the randomly generated noise is as follows:
given an initial value x0Generating a random number y over the (0,1) intervaliWherein: 2000 for a, 1 for c, M225(ii) a Then by transforming Zi＝a+(b-a)yiGenerating a random number Z over the (a, b) intervali。
3. The safety helmet identification method based on deep learning of claim 1 is mainly characterized in that: the step S3 includes the steps of:
step S3.1: the SElayer mechanism is added into native YOLOv5, namely the attention mechanism is added into BackBone network of YOLOv 5: adding an SE module into the Focus structure, and then performing information refine on the SE module and the Focus structure to improve the sensitivity of a YOLOv5 model to channel characteristics so as to improve the detection performance of the model;
step S3.2: a ghestbottleeck detection mechanism is added into native YOLOv5, namely a GHOST module is used in a Neck network layer of YOLOv5, and under the condition that the size of an output characteristic diagram is not changed, the total number of required parameters is reduced, and the calculation complexity is reduced;
step S3.3: the data set created in step S2 is used to train the yollov 5 model improved in steps S3.1 and S3.2, and when the model has good robustness, the training is finished.
4. The safety helmet identification method based on deep learning of claim 1 is mainly characterized in that: the step S4 includes the steps of:
step S4.1: and (4) applying the model trained in the step (S3.3) to safety helmet detection on a construction site, and matching 3 types of target frames person, head and helmet in the detection as follows: when a person frame appears, whether a head frame or a helmet frame appears in the person frame or not is continuously detected, and if the person frame does not appear, the head frame or the helmet frame is not detected; when a person frame is detected and a helmet frame is detected, the two frames are matched in a contact ratio manner, and when the matching degree is higher than 13%, the person is judged to wear the safety helmet; if the person box and the head box are detected, the matching degree of the person box and the head box is higher than 13%, the safety helmet is not worn, and the coincidence degree IoU is calculated by the following formula:
area (p) indicates the area of the person frame, and area (h) indicates the area of the head and the helmet frame.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111008443.1A CN113688759A (en) | 2021-08-31 | 2021-08-31 | Safety helmet identification method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111008443.1A CN113688759A (en) | 2021-08-31 | 2021-08-31 | Safety helmet identification method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113688759A true CN113688759A (en) | 2021-11-23 |
Family
ID=78584157
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111008443.1A Pending CN113688759A (en) | 2021-08-31 | 2021-08-31 | Safety helmet identification method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113688759A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113920547A (en) * | 2021-12-14 | 2022-01-11 | 成都考拉悠然科技有限公司 | Glove detection method and system based on neural network |
| CN114973080A (en) * | 2022-05-18 | 2022-08-30 | 深圳能源环保股份有限公司 | Method, device, equipment and storage medium for detecting wearing of safety helmet |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111967393A (en) * | 2020-08-18 | 2020-11-20 | 杭州师范大学 | Helmet wearing detection method based on improved YOLOv4 |
| CN112257794A (en) * | 2020-10-27 | 2021-01-22 | 东南大学 | YOLO-based lightweight target detection method |
| CN112926501A (en) * | 2021-03-23 | 2021-06-08 | 哈尔滨理工大学 | Traffic sign detection algorithm based on YOLOv5 network structure |
| CN113158956A (en) * | 2021-04-30 | 2021-07-23 | 杭州电子科技大学 | Garbage detection and identification method based on improved yolov5 network |
-
2021
- 2021-08-31 CN CN202111008443.1A patent/CN113688759A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111967393A (en) * | 2020-08-18 | 2020-11-20 | 杭州师范大学 | Helmet wearing detection method based on improved YOLOv4 |
| CN112257794A (en) * | 2020-10-27 | 2021-01-22 | 东南大学 | YOLO-based lightweight target detection method |
| CN112926501A (en) * | 2021-03-23 | 2021-06-08 | 哈尔滨理工大学 | Traffic sign detection algorithm based on YOLOv5 network structure |
| CN113158956A (en) * | 2021-04-30 | 2021-07-23 | 杭州电子科技大学 | Garbage detection and identification method based on improved yolov5 network |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113920547A (en) * | 2021-12-14 | 2022-01-11 | 成都考拉悠然科技有限公司 | Glove detection method and system based on neural network |
| CN114973080A (en) * | 2022-05-18 | 2022-08-30 | 深圳能源环保股份有限公司 | Method, device, equipment and storage medium for detecting wearing of safety helmet |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111967393B (en) | Safety helmet wearing detection method based on improved YOLOv4 | |
| CN111241959B (en) | Method for detecting personnel not wearing safety helmet through construction site video stream | |
| CN111598040B (en) | Construction worker identity recognition and safety helmet wearing detection method and system | |
| CN111144263A (en) | Construction worker high-fall accident early warning method and device | |
| CN113688759A (en) | Safety helmet identification method based on deep learning | |
| CN107437318B (en) | Visible light intelligent recognition algorithm | |
| CN110414400B (en) | Automatic detection method and system for wearing of safety helmet on construction site | |
| CN113903081A (en) | Visual identification artificial intelligence alarm method and device for images of hydraulic power plant | |
| CN113034397B (en) | High-altitude parabolic detection method capable of realizing automatic tracing of self-adaptive track in multiple environments in real time | |
| CN112364778A (en) | Power plant safety behavior information automatic detection method based on deep learning | |
| CN116259002A (en) | Human body dangerous behavior analysis method based on video | |
| CN112434827B (en) | Safety protection recognition unit in 5T operation and maintenance | |
| CN110674790A (en) | Abnormal scene processing method and system in video monitoring | |
| CN109389146A (en) | Equipment state feedback method and system based on neural network image sorting algorithm | |
| CN115035088A (en) | Helmet wearing detection method based on yolov5 and posture estimation | |
| CN113298077A (en) | Transformer substation foreign matter identification and positioning method and device based on deep learning | |
| CN112541439A (en) | Intelligent video monitoring method and system for power plant | |
| CN111695432A (en) | Artificial intelligent face abnormity detection system and method under video monitoring scene | |
| CN117496129A (en) | YOLOv 7-based improved factory safety wearing target detection method | |
| CN117612249A (en) | Underground miner dangerous behavior identification method and device based on improved OpenPose algorithm | |
| CN113554682B (en) | Target tracking-based safety helmet detection method | |
| CN115565137A (en) | Improved YOLOv 5-based unsafe behavior detection and alarm method | |
| CN114067365B (en) | Helmet wearing detection method and system based on central attention network | |
| CN114429677A (en) | Coal mine scene operation behavior safety identification and assessment method and system | |
| CN114694090A (en) | Campus abnormal behavior detection method based on improved PBAS algorithm and YOLOv5 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211123 |
|
| RJ01 | Rejection of invention patent application after publication |